AT&T is launching a pilot version of a network-based “digital receptionist” that answers unknown numbers for you, spending the time talking to the caller and determining whether the call is important enough to get access to you. Think of it as a bouncer for your phone line — looking out for fraud and chatter and letting only what matters through. The carrier states it uses speech-to-speech, high-agency AI to have a quick conversation with the caller, determine whether they are a telemarketer, and decide if it’s important for you to talk to them based on algorithms and your priorities. Early access will be available to a limited number of customers who have opted in, with the company stating that the feature is more than just a spam filter.
How AT&T’s network-based digital receptionist works
When you receive a call from a number not saved on your device, the AI answers it first. It asks very basic questions — who it is and why they are calling — and then it either lets the call through, blocks it, or sends you a message until you decide, even letting you monitor the call in real time by recording the speech. If it notices typical patterns, or if the caller declines to prove they are a person or refuses to tell you who they are, it will hang up. AT&T states that the assistant can carry out basic instructions, such as checking the designated delivery period, without having to talk to you. You can build a “Do Not Screen” list for friends who can access your number without a problem. Additionally, the carrier states that network information enables identification of regular contacts even if you don’t add them to the list.
AT&T’s AI is different from device-first capabilities because it operates in the network. No app to install, no battery penalty, no specialized service to leave running on your phone. Screening takes place on the inbound side, if at all possible — before your own phone has a chance to ring. The carrier is pitching this architecture as a way to differentiate itself from on-device call screening.
AT&T already uses technologies like STIR/SHAKEN to authenticate who’s calling, and the new receptionist is built on that system. If authentication verifies where a call came from, the AI layer essentially tacks conversational triage on top of that: an extra gate aimed at cutting out scams still slipping through.
Robocalls remain rampant despite industry-wide efforts
Even with strides being made within the industry, Americans continue to receive billions of unwanted calls each month. The YouMail Robocall Index consistently puts the total between 4 billion and 5 billion robocalls per month in the U.S., though there are seasonal peaks. In a recent press release, the USTelecom Industry Traceback Group revealed its members are cooperating to trace and shut down illegal robocalls.
The consumer stakes are real. The Federal Trade Commission says scam losses have risen into the tens of billions a year across all channels, with phone-based scams still a persistent — and very annoying — vector. And by front-stopping unknown callers and giving users prompts to force basic verification, AT&T’s assistant seeks to both decrease annoyance and discourage risk.
How AT&T’s AI receptionist compares with other tools
Smartphone platforms already offer defenses. Google’s Pixel lineup offers Call Screen, Apple has Live Voicemail and Silence Unknown Callers, and other service providers offer packages that include Scam Shield or Call Filter. These features depend on your device software, network reputation, or voicemail flow.
AT&T’s offering is a bit different: a network-native agent that speaks to callers in real time, can be programmed to perform non-complex tasks and applies user-set guidelines before your phone even starts ringing. If it functions as advertised, this can blend the best aspects of caller authentication and reputation scoring with a human-style screen.
Privacy, user control, and what AT&T says is next
Because the receptionist will answer for you, these questions about data become all the more natural. AT&T has not publicly specified the length of time it plans to keep call audio or transcripts in this trial, but privacy controls — such as the “Do Not Screen” list and real-time overrides — indicate that maintaining user agency around the tech is part of its plan. Transparency in how data is processed, stored, and used to train models will be critical for trust.
AT&T has also suggested that the assistant will grow beyond defensive screening. The company has said it is considering experiences users would “expect” and others they “may not expect.” Andy Markus, AT&T’s chief data officer, imagines that future capabilities could include booking reservations or appointments — effectively turning unknown callers into bundled tasks when you want them to be.
And for now the value proposition is simple: fewer interruptions, safer calls, and a smarter gatekeeper. Robocalls continue to be a plague and scams keep evolving; at this point, a network-level AI bouncer could well be the next significant step toward reintroducing some sanity into the phone line most of us still use.